Hematological and Biochemical Predictors of Cancer Diagnosis: An Analytical Study from a Cancer-Enriched Clinical Cohort

Abstract

Background:

Routine laboratory tests are almost universally ordered for patients investigated for possible cancer, yet it is still unclear how far these everyday measurements can help to separate malignant from non-malignant disease once a patient has reached specialist care.

Objective:

To describe the relationship between cancer diagnosis (cancer vs no cancer) and a set of simple demographic, hematological, biochemical, and clinical variables, and to examine whether any of these measures show a meaningful association with cancer status in a referred population.

Methods:

We carried out an analytical cross-sectional study of 1,000 individuals drawn from the Cancer Risk Stratification Using Lab Parameters dataset. Age, sex, smoking, family history, full blood count indices, blood glucose, CA-125, PSA, CEA, risk category, cancer stage, treatment outcome, and survival were retrieved. Cancer status (cancer vs no cancer) was the main outcome. Descriptive statistics were combined with χ² tests, independent t-tests, one-way ANOVA, and Pearson’s correlations.

Results:

Cancer was present in just over four-fifths of the cohort. Most routine blood and chemistry values lay within conventional reference limits, and for almost all variables there were no statistically significant differences between cancer and non-cancer groups. Hemoglobin was modestly but significantly lower in patients with cancer. Cancer status was strongly related to stage, and showed a weaker association with survival.

Conclusion:

In this cancer-enriched referral population, single routine laboratory measures and basic demographic factors added little to the simple knowledge of whether a patient had been staged as having cancer. Hemoglobin behaved as a non-specific marker of disease burden rather than a diagnostic test.

1. BACKGROUND

Cancer continues to be one of the leading causes of global morbidity and mortality, accounting for almost one in six deaths (1). Cancer imposes a massive clinical and economic burden on the health system (2). Further increase in global cancer burden is expected due to population ageing and ever-ongoing exposure to carcinogenic risk factors like tobacco, obesity, physical inactivity, and environmental pollutants (2). Most of the burden will be borne by low- and middle-income countries (LMICs) (2). Early detection and risk stratification are thus crucial for improving outcomes through timely diagnosis, appropriate staging, and optimized treatment decisions.

Population-based epidemiology has established several traditional risk factors, such as age, gender, smoking, and family history (3). Many studies have shown that tobacco use, family history, and older age substantially increase the chances of developing many solid and blood cancers (3–5). However, once referred to specialized oncology pathways, these factors are often prevalent in cancer and non-cancer referrals, rendering them poor discriminators in the clinical setting (4). In these rich clinical environments, clinicians increasingly rely on imaging, histopathology, and laboratory parameters to enhance their diagnostic and prognostic assessments (6).

Normal blood and chemical tests, performed daily, are inexpensive, widely available, and done on most patients undergoing cancer testing (7). The complete blood count (CBC) and basic metabolic profile may reflect systemic inflammation, nutritional status, organ function, and bone marrow reserve, all of which impact cancer biology and tolerance to therapy (7). New cancer-related anemia—eg, due to chronic inflammation, marrow invasion, nutritional deficiency, treatment toxicity—has been associated with poor quality of life and reduced survival (8). Similarly, leukocyte and platelet counts may reflect an inflammatory and prothrombotic state due to tumor growth and spread (9). It remains unclear to what extent these standard laboratory parameters may discriminate cancer from non-cancer in a referred cohort, and the current evidence is heterogeneous across tumor types and clinical scenarios (5, 6, 8).

Serum tumor markers, such as cancer antigen 125 (CA-125), prostate-specific antigen (PSA), and carcinoembryonic antigen (CEA), are often used in oncology for monitoring disease, and in some cases, screening or early diagnosis. CA-125 is an important tumor marker for epithelial ovarian cancer and plays a crucial role in its management (10). However, this marker is frequently elevated due to benign gynecologic conditions and other non-gynecologic diseases (11). This relatively common occurrence reduces the specificity of CA-125 for ovarian cancer (12). The level of CEA is commonly used in colorectal cancer for monitoring treatment response and follow-up. However, this level is also influenced by the presence of smoking and other benign conditions (13).

2. OBJECTIVE

To evaluate the association between cancer diagnosis (cancer vs. no cancer) and key demographic, clinical, and laboratory variables (age, gender, smoking status, family history, WBC, RBC, HB, PLT, blood glucose, CA-125, PSA, CEA).

3. MATERIAL AND METHODS

This study used the “Cancer Risk Stratification Using Lab Parameters” database to obtain a secondary dataset of 1000 ((). A clinical data study was performed for all patients. It included variables such as age, gender, smoking status, family history, hematological and biochemical parameters (WBC, RBC, HB, PLT, blood glucose, CA-125, PSA, CEA).

The study also included clinical outcomes (risk level, stage, diagnosis, treatment outcome, survival). The primary dependent variable was cancer status (cancer vs no cancer). The data were cleaned and coded for further analysis using descriptive statistics. The associations of categorical variables were examined with the use of χ² tests. Independent t-tests and a one-way ANOVA were used to compare groups. Pearson’s correlation was used to investigate the relationship of cancer status with continuous variables. Statistical significance was noted for p-values <0.05

4. RESULTS

4.1. Demographic and clinical features of participants

As shown in Table (1), the study included 1000 participants. The mean age of the participants was 50.2 ± 17.4 years. The gender distribution in the study was quite balanced, with 48.3% female participants and 51.7% male participants. Almost half the samples used tobacco (49.8%) while nearly half reported a family history of cancer (47.7%).

The average WBC, RBC, hemoglobin, platelets, and blood glucose were within normal limits, which indicates preserved baseline physiological status. On the other hand, tumor markers showed clinically significantly higher levels in an enriched malignancy cohort, with CA-125 (102.6 ± 58.0 U/mL) and PSA (4.97 ± 2.92 ng/mL).

From a clinical viewpoint, cancers were identified in 83.1% of the subjects, with stages I-IV being fairly evenly distributed. About 17% of the subjects had no cancer. The risk categories, low, medium, and high, were also balanced at ~31-35%. The response to treatment varied widely among patients, with quality-of-life outcomes evenly distributed across the “worst,” “bad,” “good,” and “improved” groups. At the time of follow-up, it was found that 52.2% of the patients had died.

Table 1. Demographic and clinical features of participants.

Variable	Description
Age (M±SD) years	50.20±17.37
Gender (N, %): Female Male	483 (48.3%) 517 (51.3%)
Smoking Status (N, %): Non –smoker Smoker	502 (50.2%) 498 (49.8%)
Family History (N, %): No Yes	523 (52.3%) 477 (47.7%)
WBC (M±SD)	7.05±2.04
RBC (M±SD)	4.75±0.425
HB (M±SD)	14.47±1.44
PLT (M±SD)	295.88±88.41
Blood Glucose (M±SD)	95.06±14.16
CA 125 (M±SD)	102.63±57.96
PSA (M±SD)	4.97±2.92
CEA (M±SD)	2.52±1.43
Risk Level (N, %): Low Risk Medium Risk High Risk	314 (31.4%) 347 (34.7%) 339 (33.9%)
Stage (N, %): No cancer Stage I Stage II Stage III Stage IV	169 (16.9%) 212 (21.2%) 199 (19.9%) 208 (20.8%) 212 (21.2%)
Diagnosis (N, %): No cancer Cancer	169 (16.9%) 831 (83.1%)
Treatment outcomes (N, %): Worst Bad Neutral Improved	254 (25.4%) 257 (25.7%) 258 (25.8%) 231 (23.1%)
Survival Status (N, %): Died Alive	522 (52.2%) 478 (47.8%)

4.2. The relationship between cancerous status and study variables

Table 2 examined whether the cancerous state was statistically linked to the main study variables.

Table 2. The relationship between cancerous status and study variables.

variable	Cancerous Status				P value
	No Cancer		Cancer
	N	%	N	%
Gender: Female Male	84 85	17.4% 16.4%	399 432	82.6% 83.6%	0.738
Smoking Status: Non-smoker Smoker	87 82	17.3% 16.5%	415 416	82.7% 83.5%	0.736
Family History: No Yes	94 75	18% 15.7%	429 402	82% 84.3%	0.354
Risk Level: Low Risk Medium Risk High Risk	50 60 59	15.9% 17.3% 17.4%	264 287 280	84.1% 82.7% 82.6%	0.855
Stage: No Cancer Stage I Stage II Stage III Stage IV	169 0 0 0 0	100% 0% 0% 0% 0%	0 212 199 208 212	0% 100% 100% 100% 100%	<0.001
Treatment Outcomes: Worst Bad Good Improved	50 49 35 35	19.7% 19.1% 13.6% 15.2%	204 208 223 196	80.3% 80.9% 86.4% 84.8%	0.187
Survival Status: Died Alive	100 69	19.2% 14.4%	422 409	80.8% 85.6%	0.047

Gender.

A similar pattern of distribution of males and females in having cancer or not was observed. No statistically significant difference was observed (p = 0.738).

Smoking status.

Illness occurrence is quite the same in both smokers and non-smokers (approximately 83% in both groups) with a non-significant p-value (p = 0.736).

Family history of cancer.

A slightly higher proportion of cancer cases among those with a positive family history (84.3% vs. 82.0%), but not significant (p = 0.354).

Risk level.

The proportion of cancer patients is almost equal in all low, medium, and high-risk groups (= 82-84%). The p-value (0.855) indicates no association between the predetermined risk level and the actual cancer status.

Stage.

In anticipation, the stage is appropriately aligned with the cancer status (p < 0.001). All the participants who do not have cancer belong to the “No cancer” category, and all Stage I–IV cases belong to the cancer group. If the staging variable is only applicable to cancer patients, that essentially confirms a definitional relationship.

Treatment outcomes.

Both cancer and non-cancer patients have more or less the same distribution of those who are “worst”,” bad”,” good”, or” improved”. The p-value (0.187) does not indicate a significant association between treatment outcome category and cancerous status.

Survival status.

A significant relation is present (p = 0.047). The “died” group has a higher proportion of non-cancer participants (19.2%) compared to the “alive” group (14.4%). A reverse pattern is observed for cancer patients: 85.6% of the alive are cancer patients, while only 80.8% of the dead are cancer patients. This means that the survival status has a modest but statistically significant relationship with cancer diagnosis, reflecting the mortality pattern in this mixed cohort.

4.3. The relationship between cancer diagnosis and hematological variables

According to Table 3, most of the hematological parameters among patients did not show significant differences. In both cancer and non-cancer groups, the means of WBC, RBC, PLT, blood glucose, CA-125, PSA, and CEA are similar, and the high p-values (p > 0.05) indicate no significant difference between the groups. The only statistically significant difference between them is Hemoglobin (HB) (p=0.017).

Table 3. The relationship between cancer diagnosis and hematological variables.

Variable	Diagnosis	N	Mean	Std. Deviation	P value
WBC	No Cancer	169	7.1670	1.99834	0.421
	Cancer	831	7.0282	2.05105
RBC	No Cancer	169	4.7460	.41870	0.769
	Cancer	831	4.7565	.42642
HB	No Cancer	169	14.7078	1.46365	0.017
	Cancer	831	14.4179	1.42913
PLT	No Cancer	169	301.5633	92.13268	0.360
	Cancer	831	294.7296	87.65127
Blood glucose	No Cancer	169	96.8767	14.06868	0.068
	Cancer	831	94.6964	14.16028
CA_125	No Cancer	169	97.7803	57.45387	0.231
	Cancer	831	103.6170	58.05059
PSA	No Cancer	169	4.9566	2.94378	0.933
	Cancer	831	4.9773	2.91317
CEA	No Cancer	169	2.7104	1.36738	0.60
	Cancer	831	2.4833	1.44371

4.4. Correlation between cancer status and study variables

As indicated in Table 4, the correlation between most of the study variables and cancer status was very weak and non-significant.

Table 4. Correlation between cancer status and study variables.

Variable	Correlation Level	Significance
Age	- 0.013	0.672
Gender	0.013	0.689
Smoking Status	0.012	0.716
Family History	0.030	0.343
WBC	-0.025	0.421
RBC	0.009	0.760
HB	-0.076	0.017
PLT	-0.029	0.369
Blood Glucose	-0.58	0.068
CA_125	0.030	0.233
PSA	0.003	0.933
CEA	-0.059	0.060
Risk Level	-0.016	0.618
Stage	0.675	<0.001
Treatment Outcomes	0.057	0.070
Survival Status	0.063	0.047

Only three variables displayed statistically significant correlations. There is a weak negative correlation between hemoglobin (HB) and cancer (r = −0.076, p = 0.017). There is a strong positive correlation between the stage and cancer status (r = 0.675, p < 0.001). According to the study, there is a weak but significant positive correlation between cancer diagnosis and survival outcome (r = 0.063, p = 0.047).

4.5. The impact of study variables on cancerous status

Table 5 gave an overview of differences in measurements among the cancer and non-cancer groups using the Way ANOVA test. For most factors (age, gender, smoking status, family history, WBC, RBC, platelets, blood glucose, CA-125, PSA, CEA, risk level, treatment outcomes), the F values were low and p > 0.05, which implies no significant difference between cancerous and non-cancerous participants. The result showed that hemoglobin (HB) level had a significant effect (F = 5.73, p = 0.017). The stage of cancer has the most substantial impact (F = 833.35, p < 0.001), as expected; staging indicates the presence of cancer. The survival status also has a small and significant effect (F = 3.97, p = 0.047). The p-values for blood glucose, CEA, and treatment outcomes were borderline (≈0.06–0.07).

Table 5. The impact of study variables on cancerous status.

variable		Sum of Squares	df	Mean Square	F	Sig.
Age	Between Groups	54.143	1	54.143	0.179	0.672
	Within Groups	301461.857	998	302.066
	Total	301516.000	999
Gender	Between Groups	0.040	1	0.040	0.160	0.689
	Within Groups	249.671	998	0.250
	Total	249.711	999
Smoking Status	Between Groups	0.033	1	0.033	0.133	0.716
	Within Groups	249.963	998	0.250
	Total	249.996	999
Family History	Between Groups	0.224	1	0.224	0.898	0.343
	Within Groups	249.247	998	0.250
	Total	249.471	999
WBC	Between Groups	2.706	1	2.706	0.649	0.421
	Within Groups	4162.536	998	4.171
	Total	4165.243	999
RBC	Between Groups	0.016	1	0.016	0.086	0.769
	Within Groups	180.376	998	0.181
	Total	180.392	999
HB	Between Groups	11.798	1	11.798	5.729	0.017
	Within Groups	2055.105	998	2.059
	Total	2066.902	999
PLT	Between Groups	6558.496	1	6558.496	0.839	0.360
	Within Groups	7802734.235	998	7818.371
	Total	7809292.730	999
Blood Glucose	Between Groups	667.597	1	667.597	3.337	0.068
	Within Groups	199678.063	998	200.078
	Total	200345.659	999
CA_125	Between Groups	4784.332	1	4784.332	1.425	0.233
	Within Groups	3351552.005	998	3358.269
	Total	3356336.337	999
PSA	Between Groups	0.060	1	0.060	0.007	0.933
	Within Groups	8499.683	998	8.517
	Total	8499.743	999
CEA	Between Groups	7.242	1	7.242	3.536	0.060
	Within Groups	2044.081	998	2.048
	Total	2051.323	999
Risk level	Between Groups	0.162	1	0.162	0.248	0.618
	Within Groups	652.213	998	0.654
	Total	652.375	999
Stage	Between Groups	881.550	1	881.550	833.348	<0.001
	Within Groups	1055.726	998	1.058
	Total	1937.276	999
Treatment Outcomes	Between Groups	4.018	1	4.018	3.301	0.070
	Within Groups	1214.826	998	1.217
	Total	1218.844	999
Survival Status	Between Groups	.988	1	0.988	3.969	0.047
	Within Groups	248.528	998	.0249
	Total	249.516	999

5. DISCUSSION

In this study, we revisited a fairly typical but cancer-enriched hospital cohort and asked a simple question that many clinicians recognize from daily practice: do the “standard” laboratory results and a few background characteristics really help us to distinguish patients with cancer from those without it? At first sight one might expect at least some clear separation, yet our analyses suggest that the picture is more nuanced.

More than 80% of our sample had a malignant diagnosis, which reflects the fact that these patients had already entered an oncology pathway. Within such a group, classical risk factors such as older age, male sex for some tumors, smoking, and a positive family history are common in both cancer and non-cancer patients. In our data these variables showed no meaningful association with cancer status. This does not contradict the epidemiological literature; rather, it illustrates that once people have been referred because cancer is suspected, upstream risk factors lose much of their discriminative value.

The routine hematological and biochemical parameters behaved in a similar way. With the exception of hemoglobin, values for the full blood count and basic chemistry tests were not significantly different between cancer and non-cancer groups (3). Clinicians will recognize this pattern: apart from specific situations such as acute leukemia, advanced marrow infiltration, or very aggressive systemic disease, most patients with solid tumors do not present with extreme abnormalities in their blood counts. Instead, they often sit within broad “normal” ranges that overlap substantially with those of patients who turn out not to have cancer (3).

Hemoglobin was the one variable that did show a statistically significant difference. Patients with cancer tended to have slightly lower levels than those without, and hemoglobin was negatively correlated with cancer status in the correlation analysis. This fits well with the established observation that anemia is common in malignancy and can arise through multiple mechanisms, including chronic inflammation, reduced nutritional intake, marrow suppression, and blood loss (7, 8). Nevertheless, the effect size in our cohort was small, and hemoglobin on its own would clearly be an unreliable diagnostic marker; many non-cancer conditions also lower hemoglobin, and some cancer patients maintain normal levels throughout their illness (9).

Tumor markers in the dataset (CA-125, PSA, CEA) did not, in aggregate, separate cancer from non-cancer. This is perhaps disappointing but not unexpected. Each of these markers has recognized roles in particular settings — for example, CA-125 in ovarian cancer and PSA in prostate disease — but all suffer from limited specificity and are influenced by a variety of benign conditions. In a mixed referral cohort such as ours, any signal carried by these markers is diluted by the heterogeneity of underlying pathologies. (9, 10)

One interesting finding was the lack of correspondence between the predefined “risk level” variable and actual cancer status, whereas stage and survival were, as expected, much more closely related to the presence of malignancy (11). This suggests that the way risk was originally categorized in the dataset does not capture the joint information present in the raw variables. Modern prediction-modelling approaches emphasize that risk scores should be derived from multivariable models with internal and external validation, rather than from informal groupings that have not been rigorously tested (9-12).

The weak association between survival and a simple yes/no cancer variable also merits comment. More than half of the cohort had died by follow-up, which reflects both the serious nature of cancer and the age and comorbidity profile of the population. Many deaths in older or multimorbid patients are due to competing non-cancer causes, so survival cannot be interpreted as a pure proxy for cancer burden. A proper survival analysis, which would take time and competing risks into account, was beyond the scope of this cross-sectional work but would be a logical next step (13).

Taken together, our results support a pragmatic clinical message. In a specialist oncology setting, neither routine demographic factors nor single laboratory measurements provide strong discrimination between cancer and non-cancer once patients have already been referred (15). Hemoglobin behaves as a modest, non-specific indicator of overall illness rather than as a diagnostic test, and common tumor markers perform poorly when applied in a broad, unselected way (15). For meaningful risk stratification, more sophisticated models that combine multiple markers with clinical information, and that are carefully validated and transparently reported, are likely to be required (16).

6. CONCLUSION

This cancer-enriched cohort showed that familiar risk factors and routine blood tests offer only weak discrimination between patients with and without cancer once they have entered an oncology pathway. Hemoglobin was slightly lower among cancer cases, but the difference was small and clearly insufficient for use as a stand-alone diagnostic tool. Predefined risk categories also tracked poorly with actual diagnoses, whereas stage and survival were more closely related to malignancy, as would be expected.

These findings argue against placing heavy weight on individual laboratory parameters when clinicians attempt to decide whether a referred patient does or does not have cancer. Instead, they point towards the need for multivariable, well-validated prediction models that integrate routine tests with clinical and staging information, and for careful evaluation of how such tools perform in real-world practice.

Author’s Contribution:

The all authors were involved in all steps of preparation of this article. Final proofreading was made by the first author.

Conflicts of interest:

There are no conflicts of interest.