We identified 26 studies (see Table 1) which had performed follow-up assessments of patients diagnosed with schizophrenia or related disorders. These studies were presented by 18 groups, 3 of which (29, 31, 37–39, 41–43, 47, 52, 53) had carried out two to five studies each. Two other groups published two to three papers each on different follow-up examinations of one patient samples (30, 48–50, 58). Therefore, the latter data were merged for this review’s purpose.

The number of patients included ranged from 10 to 126, with less than 20 patients in 7 studies; 15 studies dealt with FE patients. Patients suffering from bipolar disorder, major depressive disorder, affective disorders, delusional disorder, substance-induced psychotic disorders, and personality disorders accounted for 21 and 36% of participants in two studies (44, 58); all remaining exclusively consisted of subjects suffering from schizophrenia.

Instruments for the assessment of NSS were mostly standardized and widely used scales, namely the Neurological Evaluation Scale [NES (62)], the Cambridge Neurological Inventory [CNI (61)], the Condensed Neurological Examination [CNE (60)], the Heidelberg Scale [HD (29)], and the Brief Motor Scale [BMI (59)]. The following six exceptions were made: Emsley et al. (41, 42) shortened the NES to 13 items, Madsen et al. (44) employed a standard neurological examination complemented by tests of sensory function and complex motor acts, Mittal et al. (46) used the Quitkin scale (2), Tucker and Silberfarb (1) applied 4, Torrey (56) and Wahlheim et al. (57) two single tests respectively.

Apart from adjustments and exceptions, the different NSS scales and scores are not comparable directly since the number of items differs between instruments as does the scoring (0–1 to 0–3 points) of the individual items. However, items and/or subscales are contrastable, especially with respect to motor sequencing, motor coordination, and sensory integration, which are the most relevant items concerning pathology in schizophrenia. These important components of NSS were assessed in almost all studies integrated in this review.

Ratings were performed by 2–3 raters in half of the studies, achieving overall good interrater reliabilities (see Table 1). One single rater applied the tests in the remaining 50%, which was mainly the case in older studies.

Catamnestic periods extended from a short-term of 2–8 weeks to a medium duration of 6 months to 2 years to long-term follow-ups of 5 and 6 years (see Table 1).

Researchers who included HC in their assessments unanimously reported that patients scored significantly worse than healthy comparison subjects [review by Dazzan and Murray (6)]. Table 1 indicates that 35% of all studies in this review compared to HC patients, all of which showed significant differences in NSS scores throughout.

For reasons of better comparison, we grouped all identified studies according to patients’ illness status and their NSS course (Table 2). Independent of the number of patients included and the instruments used for the assessment of NSS, all studies on chronic schizophrenia showed that NSS remained stable or deteriorated, whereas studies on patients with a remitting course found that NSS decreased over time. The pattern also appears in the studies by Schröder et al. (29, 52, 53), in which both chronic and remitting patients participated. There are two exceptions, namely the studies by Jahn et al. (43) and Torrey (56) where mixed samples were included and the respective NSS results on follow-up are heterogeneous. Presumably the divide follows the above model, which, however, cannot be proven or rebuted.

The picture on FE patients is less clear. Whereas patients’ NSS status improved in several studies (30, 31, 42, 45, 58), the patients’ cohort fell into two subgroups in the majority of studies, mostly with one subgroup exhibiting an improving and another a stable NSS course over time.

Only Boks et al. (35) and Madsen et al. (44) reported a division into two subgroups with stable and deteriorating NSS, and Prikryl et al. (48–50) described two subgroups with improving and worsening course, respectively. No change except for worsening of the glabellar reflex was observed in one study (51).

Change in total NSS—if present—was mostly accounted for by change in the motor system subscales, i.e., motor coordination and motor sequencing (29, 31, 38, 49, 51, 58). Cuesta et al. (40) reported an improvement in all subscales except for frontal signs.

Emsley et al. (41) reported a decrease of motor signs during the first 3 months of treatment with a subsequent return to initial values. Accordingly, in their second study, the group (42) did not note a drop in motor signs despite of a significant decline in total NSS. Other authors divided their samples into remitters and non-remitters and reported a decrease of motor signs being related to a remitting course (50 plus sensory integration; 29 plus spatial orientation). Likewise, in a FE study (31) subgroups arose after 1 year, which were distinguishable by the subscales motor coordination, motor sequencing, and sensory integration. The sensory integration subscale also added to an overall change in NSS scores in two other studies (51, 38), right-left and spatial orientation did so in one (31). Moreover, one group detected a non-significant decrease of the above-indicated subscale scores (54) and another group described a concordant finding for sensory integration (34).

Among the groups that did not detect any NSS change over time, Smith et al. (55) pointed out the relatively high stability of motor signs as opposed to the low stability of sensory integration signs in chronic patients. In FE individuals two groups found stability of motor signs (39, 47).

Chen et al. (38) reported deterioration of motor coordination, sensory integration, and disinhibition in a complete sample of chronic patients. Unfortunately, Boks et al. (35), Mittal et al. (46), and Whitty et al. (30) did not give any information on subscales when there was an overall NSS deterioration. In the study by Madsen et al. (44), NSS increase was mostly related to corticospinal tract signs.

Medication influences both clinical variables and NSS (see below). Nevertheless, the relationship between NSS and clinical variables will be discussed separately.

Two studies did not assess or indicate psychopathology. The Scale for the Assessment of Negative Symptoms, and the Scale for the Assessment of Positive Symptoms were used by two groups, completed by the Present State Examination (PSE-9) once, whereas all remaining researchers employed the Positive and Negative Syndrome Scale (PANSS) or the Brief Psychiatric Rating Scale (BPRS). The respective scores at study inception and follow-up are given by 13 authors, whereas, 6 reported incomplete data (see Table 1). Scores are essentially comparable in the sense that (a) scores assessed with the same instrument fall into the same range, (b) the absolute follow-up scores are even similar, (c) acutely ill patients’ score decrease over time, and (d) scores of chronic patients or of those who were assessed in a stable condition remain widely unchanged.

Several authors have reported an improvement of NSS in parallel to a decrease of symptoms or treatment response, respectively (1, 29–31, 40, 42–44, 46–52, 54, 56–58). This pattern was even found in patients without symptom change (44) and in a subgroup without negative symptoms (34). Furthermore, associations of remission and motor signs were described (44, 58, 30), whereas remitters in the studies by Prikryl et al. (48–50) exhibited lower total, sensory integration, and motor sequencing scores. Also severity of illness and lower social functioning (44) or severity of illness, except for motor sequencing, correlated with NSS (47). Other researchers did not find an association between NSS with symptoms in first-episode (35, 45) or chronic patients (36, 55). In two studies, even a deterioration of NSS in chronically ill (38) and of corticospinal functions in non-remitting FE patients was detected (44).

Positive symptoms were only addressed separately in a few studies. Scheffer (51) reported significant associations of NSS change in motor coordination, sequencing of complex motor acts, and sensory integration with change in the positive subscale (and total BPRS score, as discussed above). This is supported by findings of significant positive correlations between NSS and positive symptoms in two studies (45, 46) and of balance and positive symptoms in another study (41). Results by Chen et al. (38) did not underline the aforementioned findings. Even after a split into high and low positive symptoms no association with NSS emerged.

Researchers reported more extensively on negative symptoms compared to other symptoms present at the catamnestic examination. Many groups found a positive or increasing correlation over time between total NSS scores and negative symptoms (30, 37, 40, 42, 44, 48–53, 58), as well as correlations between motor signs and negative symptoms (39, 47). One group described partial correlations of negative symptoms with primitive reflexes (34). Yet others reported an association of NSS with discomfort (46) or with negative and disorganized symptoms (47). Some groups depicted constancy of negative symptoms and NSS over time (34, 54, 55) in diverse patient samples and even in individuals with variations of NSS. Whitty et al. (30) even posit that negative symptoms rather than positive or general psychopathological symptoms may predict NSS.

Some authors considered further symptoms in their reports. They found an association between rapid movement or convergence and depression or anxiety (41), of total NSS with disorganization symptoms (42) on the one hand and a negative correlation of NSS and discomfort (46) on the other hand.

The meta-analysis by Chan et al. (16) already reported a strong relationship between NSS and symptoms in general. This notion is confirmed and extended by longitudinal studies which argue that the association even increases over time, the longer the more, and holds true mostly for negative symptoms.

Results on outcome are thus divergent: authors report (a) either no difference between good and poor outcome patients with respect to motor signs (39) or (b) better outcome measures in patients who complied with treatment and exhibited decreasing NSS scores (31); along these lines, remission was inversely correlated to NSS as described by two groups (29, 48, 49). In terms of prediction, bidirectional relationships are possible: NSS predict symptomatology (46), positive and negative symptoms, as well as other parameters predict increasing NSS (30). From the perspective of outcome, there seems to be a close relationship to NSS and psychopathology, in the sense that both higher NSS and higher psychopathology are related to a worse outcome.

Cognitive functioning was addressed in several studies. Chan et al. (37) did not find any group*time effect with NSS and overall cognitive parameters. However, over time, NSS correlated with poorer performance on the letter–number span test. Their group also reported paralleling trends of NSS and Winsconsin Card Sorting Test (WCST) perseverative errors in both patient groups as well as higher scores in logical memory, delayed logical memory, and WCST category, which authors relate to worse motor coordination and total NSS scores. In another study (42), working memory amelioration was predicted by sensory integration and motor coordination, whereas motor sequencing was predicted by working memory. Jahn et al. (43) even found correlations between NSS and all subscales of the modified WCST. Another study from Germany (52) reported a parallelizing improvement of both NSS and cognitive parameters, namely d2, Tower of Toronto test, SKT: delayed recall (Syndrom-Kurzztest, short assessment of cognitive performance), and SKT: delayed recognition. Further nonsignificant improvements occurred in the Tower of London test and SKT: immediate recall. Along these lines worsening of higher cognitive functioning paralleled higher hard signs (30, 46), and patients with NSS—as opposed to those without—exhibited negative symptoms and cognitive disorganization (57). Respective correlations emerged in cross-sectional research (64–68).

Patients included in the reviewed study were of relatively young age. None of the authors discussed age with respect to NSS. The same holds true for gender. The literature does not report any influence of sex on NSS. Neither did we detect any influence. Family history was addressed by Madsen et al. (44) who found an NSS deterioration in a subgroup with first-degree relatives who suffered from a psychiatric disease. The same authors reported a worsening of NSS in patients with a history of obstetric complications. A relationship between level of education and NSS on follow-up was reported by two groups (39, 58, 30).

Handedness was examined where indicated in Table 1. Whereas our group did not detect any influence of handedness on NSS, Whitty et al. (58, 30) found higher NSS in mixed handers. The latter may be related to a common basis of both, mixed-handedness and NSS, consisting of still ill-defined neurological abnormalities.

Several authors addressed the duration of untreated psychosis (DUP) which was positively associated with NSS (39, 56, 58, 30) and negative symptoms or with motor symptoms only (39, 41). Similar associations arose with the overall duration of illness (45, 46, 55). Whitty et al. (58, 30) even posit that the DUP predicts catamnestic levels of NSS.

Alcohol and substance use were exclusion criteria in almost all studies, thus making the study population less representative. When this was not the case, alcohol did not influence NSS (57) or, together with negative symptoms, predicted higher NSS scores (58, 30). The latter finding is supported by the literature (69), stating that drug and alcohol abuse are associated with more neurological abnormalities in schizophrenia.

Patients were treated with typical neuroleptics in 10 studies, with atypical compounds in 5 studies, and with any of both or a mixture in the remaining 11 studies, where authors not always gave the respective information or, e.g., just displayed chlorpromazine-equivalents.

Five groups reported on a sample of antipsychotic-naïve or about 90% medication-naïve patients; another two groups included antipsychotic-naïve individuals but to a lesser extent (51, 29). A washout phase prior to starting the study was reported by two researchers (46, 54).

Dosages, where given, are not comparable between studies due to the different ways of reporting. Cuesta et al. (40), who used different antipsychotics or combinations, explicitly stated that there was significant NSS improvement over time independent of the type of medication, whereas Smith et al. (55) observed a trend toward lower NSS scores on atypical antipsychotics.

Whether or not antipsychotic medication exerts an effect on NSS can neither be proven nor disproven because the effect can only be generated indirectly. Although there is no indication regarding different effects of typical versus atypical compounds, evidence from the reviewed studies points toward different medication effects: (a)

a positive influence of medication on NSS in terms of a decrease mostly stems from FE patients (1, 31, 40, 45, 51, 58), but also from those with a longer-standing illness (46, 54, 55). Others found that medication intake was associated with less motor signs as opposed to a non-medicated status (30, 35, 58).

In summary, medicated patients seem to fall into at least two subgroups, namely one which improves clinically and in terms of NSS, and another which exhibits stable and/or worsening psychopathology as well as NSS. Possibly, the effect of medication may be related to the stage of illness and the overall magnitude of NSS which have been said to be predictors toward response to medication (70). Despite the diverse results, there is thus room for supporting the notion that antipsychotic compounds exert a protective effect on NSS (6). Already, Heinrichs and Buchanan (14) argued similarly by reporting that medicated patients exhibited fewer signs.

Consequentially, the relationship between clinical response to medication and NSS improvement—or non-improvement, respectively—leads to the assumption of a common underlying factor. For example, this may be the effect of antipsychotics on brain structure as reviewed by Scherk and Falkai (71). Ultimately though, the link between medication, brain structure, and NSS has not been located yet.

Side-effects of medication were assessed by 15 groups, using widely accepted instruments. Six authors indicated that no SE were present, five others did not give any results, hereby suggesting that non-reporting is due to a lack of significant pathology. A minority of six groups reported associations between NSS and extrapyramidal side-effects: a positive correlation on follow-up (47), a relationship to motor signs (44), to motor signs on 2-year follow-up (41), to sensory integration (45), or to stable, i.e., higher NSS (57). A significant increase of tardive dyskinesia on follow-up was observed in haloperidole treatment (36).

The presented follow-up studies on NSS in schizophrenia were performed in different countries on four continents, namely, (North) America, Europe, Asia, and Africa. So far, study results do not show any pattern in favoring or contradicting an association of NSS and ethnicity, although more studies are warranted. For now, there is reason to assume that NSS may be remarkably similar across countries and ethnicities (72).

Representativeness of the reviewed studies is high for the following reasons: only two research groups included diagnostic groups other than schizophrenias, the majority of researchers followed patients for at least 6 months, 47% of the studies focused on FE patients, and interrater reliability or intraclass correlation was high overall. Different assessment instruments do not seem to exert an influence: although different researchers have used different instruments, these were standardized and featured the same dimensions of NSS. Moreover, there is agreement in the literature on the relative importance of the different groups of NSS: motor signs, i.e., motor sequencing, complex motor tasks, and motor integration are of overriding importance as they usually account for the largest part of the total NSS score. Sensory integration on the other hand plays a minor role.

As schizophrenia has a genetic basis in up to 50% of cases (9), NSS have been looked at as endophenotypes, i.e., markers which help to discover the underlying genetic basis. There is an assumption that the amount of NSS and the abnormality of the single sign may even discriminate sporadic and familial schizophrenia with more pronounced abnormalities in familial cases (51). Along these genetic lines, many studies have reported more NSS and higher item scores in patients’ relatives compared to HC; the closer the relative, the more explicit the abnormalities (21). On the other hand, according to Kinney et al. (73) and Rossi et al. (60) NSS have a small to modest share in the liability toward schizophrenia. Gureje Gureje (74) even stated that NSS might not be specific to schizophrenia but related to obstetrical complications in general. Madsen et al. (44) may reconcile the opposite views: in their study differences between patients and healthy subjects regarding frontal, cerebrospinal, and temporo-parietal functions became larger over time—which was more prominent in patients with a positive family history. Similarly, Hyde et al. (75) claim that “frontal release signs” are influenced by both genetic and environmental factors.

Meehl (76) stated that NSS can only be considered to represent genetic markers if:

they are specific for schizophrenia, i.e., discriminate patients from HC and patients with other psychiatric diseases;

Meehl’s first requirement for a genetic marker partially holds true in NSS. There was congruence among the reviewed studies that patients exhibit more NSS than HC. This is supported by the literature including FE patients (6, 77). However, NSS can also be detected in patients suffering from a number of other psychiatric diseases including bipolar disorder, obsessive–compulsive disorder, borderline personality disorder (78–80).

Meehl’s second postulation is fulfilled. NSS were also detected in siblings, but were qualitatively and quantitatively less distinct, i.e., taking an intermediate position between HC and patients. This represents a strong argument for a trait or hereditary component as backed up by the literature on patients’ first-degree relatives (20), on twins (21–24), and on off-spring (25–27).

Requirement three demands comparability between acutely ill and remitted patients. Some studies in our review argue in favor of comparability by reporting completely stable NSS over time. This was the case even in four FE studies (35, 39, 41, 51) and in studies on chronic patients (36, 39, 55). The other studies reported differences between subgroups. However, the above-indicated FE patients presented with a long duration of untreated illness and might already have crossed the border toward chronicity, e.g., in the study by Chen et al. (39) where DUP amounted to 474 ± 768 days. Along these lines, some authors suggested an interrelationship between NSS and duration of illness (81, 83, 84). Moreover, there was one report (38) on a clear deterioration of NSS in chronic patients and a further one on patients worsening on a trend level (44), related to a non-remitting course, a family history of psychotic disorder in a first-degree relative, obstetric complications, and male gender.

Further arguments in favor of non-comparability between acutely ill, remitted, and chronically ill patients arise from studies which showed NSS improvement over time in most or all patients, differences in phases of the disease and other varying factors notwithstanding. Several follow-up studies found NSS improvement, especially in FE patients (30, 31, 42, 45, 49, 58) with respect to subscales in FE (16, 34), in patients with a remitting course (1, 29, 44, 49, 56, 57) with significant effects in two of the cited studies, and in patients who were unmedicated prior to study inclusion (46, 54). Moreover, in the study by Sevincok and Topaloglu (54), patients who were unmedicated at first assessment and medicated at the second displayed an absolute decrease of NSS—possibly due to the small number of patients the results did not reach statistical significance. In all probability, compliance plays a role toward improvement (31). As adherence most likely is particularly high in patients who participate in a study, compliance presumably was present in the presented studies. Thus, at least at the onset of the illness and in remitting subtypes an NSS state component is present and expressed by fluctuations of NSS scores. Given the number of studies reporting fluctuating and decreasing NSS in FE and remitting patients, and the fact that only chronic patients experience stability or deterioration of NSS, the state component cannot be ignored. The fact that patients’ samples mostly divide into those with stable NSS and improving NSS at an early stage during their course of illness, is in line with schizophrenia not just representing one homogeneous disease but a group of entities, namely the “group of schizophrenias” as Bleuler put it (82). As the existence of subgroups is not debatable (83, 85)—subtypes may represent another argument for the existence of an NSS state feature in some patients but not others. Unfortunately, subtypes of the disorder and their influence over time were discussed by Schröder et al. (29) only.

The fourth claim by Meehl states that schizophrenia patients with varying degrees of severity should not differ regarding NSS. Severity of patients’ illness in the different studies can either be compared by authors’ classification as FE, remitting, and chronic course or by looking at psychopathology. Whenever scores of well-known scales such as BPRS or PANSS were disclosed, their range was comparable. The latter comparison seems more valid because the former already is based on symptoms. With respect to psychopathology and NSS, differing patterns arise in those studies, which gave all necessary information. Some authors report a paralleling course of NSS and psychopathology in all patients (30, 31, 40, 42, 46, 54, 58) or in subgroups (44, 48–50, 52). Others describe decreasing symptoms (except negative symptoms) but stable NSS (34) or the opposite, namely stable symptoms and decreasing (34, 53) respectively increasing NSS (38). Yet, further studies found unchanged symptoms as well as NSS over time (35, 36).

With this diffuse pattern, Meehl’s fourth demand is clearly not fulfilled. However, the pattern speaks in favor of a state component in NSS and the necessity to consider the stage of illness, as Smith et al. (55) reasoned: if a state dimension was predominant, the timing of NSS assessment would be important. This argument also calls into question the predictive value of NSS which was proposed by earlier authors (66, 87).

Thus, only two of four of Meehl’s claims are fulfilled and NSS qualify for both, a stable, and early-acquired trait and a fluctuating state component. The latter may possibly be related to the stage of the disease as well as to treatment.